Collection of Data
Data are numerical facts obtained by counting or enumerating. Before collecting data the investigator must be clear about the objective, scope, source, method, units, accuracy required, and type of enquiry. There are two types of data — primary (collected for the first time by the investigator, original) and secondary (collected by someone else and used by another person). Five methods of primary data collection exist: direct personal interview, indirect oral interview, information from correspondents, mailed questionnaire, and questionnaire sent through enumerators — each with merits and demerits. Sources of secondary data are published (government reports, journals, books) and unpublished (research, records). Precautions are needed because secondary data may be biased, outdated, or unsuitable.
In this chapter
Data — Meaning & Definition
Data are numerical facts obtained by counting (e.g. number of students in a class) or measuring (e.g. height of students in cm). The word comes from the Latin datum (singular, "something given") — its plural is data. For example, the Central Bureau of Statistics (CBS) collected data on every household in Nepal during the National Population and Housing Census 2021, recording 29,164,578 people. Raw data are like raw rice — useful only after processing (cleaning, organising, analysing). Before collecting data the investigator must be clear about several pre-requisites so that time, money and effort are not wasted.
Pre-requisites of Data Collection
- Objective — clearly state why data are being collected (e.g. CBS collects Census data to plan schools, roads, seats in Parliament).
- Scope — define what to include and exclude (e.g. Census 2021 covered all residents of Nepal, excluding foreign diplomats).
- Source — decide whether to use primary (collect fresh) or secondary (use existing) data.
- Method — choose the appropriate technique (interview, questionnaire, observation etc.).
- Units — define the unit of measurement (e.g. income in Rs/month; age in completed years; land in ropani or hectare).
- Accuracy / Reasonable accuracy — decide how exact the data must be (Census needs 100% coverage; sample survey allows ±3% error).
- Type of enquiry — census (all units) vs sample (representative subset); official vs private; regular vs ad-hoc.
Primary vs Secondary Data
Primary data are collected for the first time by the investigator for a specific purpose — they are original. Example: a CBS enumerator going door-to-door in 2021 to ask each household about its size, age composition, and literacy is collecting primary data. Secondary data are collected by someone else for their own purpose, and later used by another investigator. Example: if a researcher downloads the Census 2021 report from the CBS website and uses it to study provincial migration patterns, those data are secondary for the researcher. Both have their place — primary data are accurate but costly; secondary data are cheap but may be outdated or not exactly suited to your question.
Primary data vs Secondary data — comparison
| Basis | Primary Data | Secondary Data |
|---|---|---|
| Originality | Collected for the first time, original | Already collected by someone else |
| Collector & user | Same person / agency | Different — collector ≠ user |
| Cost | High (time, money, manpower) | Low (often free, just download) |
| Time | Long — months or years | Short — hours or days |
| Accuracy / Suitability | High — exactly fits the question | May be outdated or not exactly suited |
| Example (Nepal) | CBS Census 2021 enumeration | Researcher downloading Census 2021 report |
| Reliability | High — investigator controls quality | Depends on the original collector |
Methods of Primary Data Collection
- Direct personal interview — the investigator meets respondents face-to-face and asks questions (e.g. a teacher surveying Class 12 students about study habits).
- Indirect oral interview — instead of asking the respondent directly, the investigator questions third parties who know the respondent (used when the topic is sensitive, e.g. asking neighbours about a family's income).
- Information through correspondents — local agents (correspondents) in different places collect and send data periodically (e.g. NRB uses correspondents in markets to collect weekly price data for CPI).
- Mailed questionnaire — a questionnaire is posted or emailed to respondents who fill and return it (e.g. a researcher emailing a questionnaire to all 753 municipalities).
- Questionnaire sent through enumerators** — trained enumerators carry the questionnaire to respondents, explain questions, and fill in the answers (e.g. CBS Census 2021 used 40,000 enumerators)
Five methods of primary data collection — merits & demerits
| Method | Merits | Demerits | Nepal Example |
|---|---|---|---|
| Direct personal interview | High accuracy; flexible; can clarify questions; high response | Time-consuming; costly; small area; personal bias possible | Teacher surveying 50 students in their own school |
| Indirect oral interview | Wide area covered; less cost than direct; useful for sensitive topics | Less accuracy (third-party info); respondent may be unwilling | NGO asking neighbours about a poor family's income for aid |
| Information from correspondents | Wide coverage; regular updates; economical for repeated surveys | Bias possible; correspondents' quality varies; no control | NRB market correspondents collecting weekly prices for CPI |
| Mailed questionnaire | Wide area; economical; quick; respondents answer at convenience | Low response rate (often <30%); only literate respondents | Researcher emailing questionnaire to all 753 municipalities |
| Questionnaire via enumerators | Wide coverage; accurate; can cover illiterate respondents; explanation possible | Very costly; needs trained enumerators; time-consuming | CBS Census 2021 with 40,000 trained enumerators going door-to-door |
The choice of method depends on (i) objective, (ii) budget, (iii) time, (iv) literacy of respondents, and (v) geographical area. For Nepal, the enumerator method is by far the most reliable for nation-wide surveys — that is why CBS uses it for every Census and Nepal Living Standards Survey (NLSS). The mailed questionnaire is rarely used because literacy and postal reliability are limited. The correspondent method is ideal for regular monitoring like NRB's price collection for CPI (Nepal has 33 market price collection centres). The direct interview is best for small academic projects.
Sources of secondary data — published and unpublished
| Type | Source | Nepal Example |
|---|---|---|
| Published — Government | Central Bureau of Statistics (CBS) reports | Census 2021; NLSS 2010/11; Statistical Year Book |
| Published — Government | Ministry / Department reports | Economic Survey (MoF); Education Statistics (CEHRD) |
| Published — International | World Bank, IMF, UNDP, ADB reports | World Development Indicators; Human Development Report |
| Published — Trade & business | Chambers of commerce, trade journals | FNCCI economic reports; Tea & Coffee Development Board |
| Published — Academic | Research journals, books, newspapers | Tribhuvan University journals; The Kathmandu Post |
| Unpublished | University theses, research projects | Master's / PhD theses at TU, KU, Purbanchal University |
| Unpublished | Government / private records | Hospital records; school enrolment registers; bank records |
| Unpublished | Internet sources | Nepaldata.com; HRD updates; blogs (verify authenticity!) |
Precautions in Using Secondary Data
- Suitability — check whether the data fit your research question (definitions, units, coverage).
- Reliability — verify the source's reputation (CBS, World Bank > a random blog); check methodology.
- Adequacy — ensure the data cover enough time period, areas, and units for your analysis.
- Definition of terms — confirm that "unemployed", "below poverty line", etc. are defined the same way.
- Unit of measurement — check that units match (e.g. income in Rs/month vs Rs/year; land in ropani vs hectare).
- Time period — make sure data are recent enough — Census 2011 is too old to study today's migration.
- Method of collection — understand how the original data were collected (sample vs census, questionnaire vs interview).
- Editing & accuracy — watch for transcription errors, missing values, suspicious outliers.
Sample size formula — used to design a questionnaire-based survey (z = 1.96 for 95% confidence; p = 0.5 if unknown; e = margin of error)
Response rate — key indicator of mailed questionnaire effectiveness
Designing a Good Questionnaire
A good questionnaire is short, simple, neutral, and logically ordered. Keep these rules: (1) introduce yourself — purpose, confidentiality; (2) start with easy, non-threatening questions (age, education) before sensitive ones (income, political views); (3) use closed-ended questions (Yes/No, multiple choice) for quick analysis, but include a few open-ended ones for depth; (4) avoid leading questions — instead of "Don't you agree agriculture is bad?" ask "How would you rate agriculture's current state?"; (5) pre-test the questionnaire on 5-10 people and revise; (6) thank the respondent at the end. CBS tested the Census 2021 questionnaire in a pilot in 2020 before the real count.
Practice Problem
You have been asked by the National Planning Commission (NPC) to design a questionnaire (about 8-10 questions) to study the living standard of households in Kirtipur Municipality. The questionnaire must include: (a) an introductory note, (b) identification questions, (c) both closed and open-ended questions, (d) at least one question on income, education, and employment each. Design the questionnaire and explain which method of primary data collection you would use and why.
Practice Problem
Identify whether each of the following is primary or secondary data for the stated investigator, and give one reason: (a) A Class 12 student surveys 30 households in their tole to study internet usage — uses the data himself. (b) A researcher downloads Nepal Living Standards Survey (NLSS) 2010/11 data from the CBS website to study poverty. (c) The Central Bureau of Statistics (CBS) collects Census 2021 data door-to-door using 40,000 enumerators. (d) A journalist quotes the Economic Survey 2023/24 figure on GDP growth in their newspaper article. (e) NRB conducts its own Dairy Market Survey in 33 markets every month for CPI computation. (f) A PhD student at Tribhuvan University uses hospital records (not their own) on dengue cases in 2022. Also compute: in part (a), if the student distributed questionnaires to 30 households and got 21 usable responses, what is the response rate?